I've been fiddling with my kernel, and have compiled one with and one without the -ck patch, using the most recent release kernel sources. I think that for light desktop use, with XFCE, the -ck patch shows better seat of the pants feel, for my light desktop use.

Does anyone know if there are any real benchmarks with these different schedulers? Files seem to delete faster (JFS) with -ck than without, although again, seat of pants.

Any downsides to other schedulers?

Does anyone use the staircase scheduler, or is it not a popular choice?

You could do write and read benchmark yourself, with

Write test :
Read test :
do these test with your different kernel or maybe use hdparm...

I was only providing the deletion of files as an example, I was more interested in any microbenchmarks showing the difference between the two on completely cpu bound operations.

Actually, in truth, even that would not verify whether I am seeing a placebo effect nearly as much as if someone chimed in that they'd seen similar improvements.

But thank you for replying, it seems that this forum is more for rescueing unfortunates with problems than lucky folks who get away with stuff like patching the kernel, etc.

You could really do with VM'ing 2 kernels side by side and throwing the same (many many!) operations at them to see how the schedulers react step-by-step and which completes fastest. I think that's about the only way you'd get a positive resolution from using a newer scheduler.

I suppose if you were feeling particularly brave you could write test harnesses (probably in C) that use the schedulers to come up with a hypothetical "better" scheduler. Hell, if you were really bored you could do it by hand on some scrap paper!

I don't know enough about the task queues to even suggest which one should be better!

I think you may well be correct in that!

Thanks for sharing the good news =)

- Piete.